Cancer is a complex disease, resulting from the combined influence of genetic susceptibilities and environmental exposures. Except in extreme cases, the environmental contribution is obscure, making the identification of genetic risk factors difficult. Hence, despite much effort, little is known about common low penetrance cancer-predisposing genetic variants. The Childhood Cancer Survivor Study (CCSS) is an ongoing longitudinal study of about 14,000 survivors of childhood cancer treated between 1970-1986. Of these, 805 have developed second malignant neoplasms (SMNs), a late complication of prior chemotherapy, of which the largest cohort is individuals treated for Hodgkin's disease (HD). As these SMNs are caused by the DNA-damaging agents used to treat cancer, we hypothesized that since the inducing environmental exposure is shared by those who do and those who do not develop the SMN, but only a subset of all patients treated develops a SMN, these individuals are genetically predisposed towards SMNs. To test this, we propose in this application to undertake a genome-wide association (GWA) study to map genomic loci harboring the genetic determinants of SMNs after HD by linkage disequilibrium (LD). 500,000 single nucleotide polymorphisms (SNPs) located throughout the genome at an average intermarker distance of 2.5 kb will be genotyped in all 128 Caucasian children followed in the CCSS treated for HD who developed an SMN, as well as an equal number of children treated for HD matched for ethnicity, race, age, and treatment who did not. To discover SMN- associated variants, genotype frequencies will be compared, and p values assigned by Fisher's exact test. By permutation of cases and controls, the false discovery rate will calculated, to allow the proportion of false positives to be estimated for any p value, thereby allowing the proportion likely to be true susceptibility alleles to be readily determined. We will also undertake a novel multi-locus analysis to compare the extent and patterns of LD in cases to controls. These two intersecting approaches will be used in concert to identify candidate susceptibility loci. Because we will have access to lymphoblastoid cell lines (LCLs) derived from each patient, follow-up studies guided by these genetic data will be undertaken to study functionally the prognostic significance of the biomarkers we identify. In a pilot study to discover susceptibility loci for therapy-induced acute myelogenous leukemia (t-AML), a particularly lethal SMN, we genotyped only 6600 markers, and successfully identified 15 candidate prognostic biomarkers, as well as a number of plausible genes likely to contain other prognostic biomarkers. Thus, these proposed studies are very likely to yield prognostic biomarkers by which cancer patients at high risk for SMNs can be identified and their treatment altered to minimize this risk. Further, because SMNs are a unique model for gene-environment interactions that may underlie most sporadic cancers, these data may provide insight into the genetic contribution to many cancers, and may also point the way towards novel molecular targets for new anticancer agents. Cancer results from the combined influences of multiple genetic susceptibilities and environmental exposures. Because it is often difficult to identify the environmental contribution to carcinogenesis, the discovery of genetic prognostic factors has been difficult. Since prior chemotherapy is the environmental contribution to second cancers, we will overcome this problem by looking for genetic differences between children who developed second cancers as compared to those who did not. [unreadable] [unreadable] [unreadable]